Collatz conjecture : The stochastic model of collatz sequences
This is the same contents as previous articles written in Japanese.
This article is NOT the proof of collatz conjecture, I constructed stochastic model of stopping times using Brownian motin model with constant drift.
Stopping times means operation times until natural number reaches 1.
We define .
If is odd, the operation times increase by 2, if is even operation times increase 1.
We assume the probability that f(n) and g(n) are even or odd is .
The variabe , "x" moves positive direction about by probability 1/2, negative direction by probability 1/2.
First, we think about time "t" that "x" reaches "0" for the first time(first passage time).
We multiply scale factor ,
(Amount of positive direction movement)-(Amount of negative direction movement) =2 by f() or g() operation.
We define the rescaled variabe, and ommite "(n)" of .
We approximate the movement of as Brownian motion with constant drift.Because drifts negative direction , and moves between .
We solve the first passage time of in Brownian motion with negative constant drift.
Second, we derive the relation between the first passage time time "t " and stopping times .
We define the movement times to positive direction as , the negative movement times as .
We deform
includes 2 steps, stopping times meet
eq.(1)
"t" confirms to inverse Gussian distribution, probability density function(PDF) is
.
https://en.m.wikipedia.org/wiki/Inverse_Gaussian_distribution
The expected value and variance of inverse Gussian distribution are
, , and by using epuation (1), we derive the formula,
We substitue epuation (1) to PDF and deform measure
, the PDF of stopping times is
This epuation means the probability that stopping times is included in is
.
We sum from 1 to , we derive the frequency distribution of stopping times in
eq.(2)
Next,we derive the formula of the maximum value of stopping times in by using the frequency distribution function.stopping times.
We define the maximum value of stopping times as .
In epuation (2), we convert ,
eq.(3)
First, we perform integration.
Considering and , we find is learger than .
Then, we expand the primary term of "u" for integrand,
We define , we represent the integral,
We consider about the case is learger than , and considering , we neglect 2nd and 3rd term.
Since the value of equation (3) at the maximum stopping times is O(1),
We expand at ,
Therefore,
Finaly, we consider about the maximum number of Collatz sequences in .
We define this number as .
In the Brownian motion with negative constant drift, the probability arriving Y>X from
is
We find the sum of from 1 to , the sum result represents the expected number that reaches starting from .
We deform the measure as
, then
eq(4)
If equals to the maximum number of Collatz sequences in , the result of eq(4) is .
Considering , we derive the formula,